Carrier roller running state health assessment method and system based on data driving

Abstract

The invention belongs to the technical field of health monitoring of mechanical rotating equipment, and particularly relates to a carrier roller operation state health assessment method and system based on data driving, and the method specifically comprises the steps: installing a vibration data sensing system near a supporting frame in the real-time operation process of a carrier roller, and completing the collection of needed data; s2, feature extraction is carried out on the sampling signal y (t) n obtained in the step S1 by adopting a time-frequency domain feature extraction method; forming a health characteristic index system matrix; the calculation of all index parameters for health assessment of the running state of the carrier roller is completed, and the obtained health index characterization is obtained; calculating the weight of a single index in the index parameter system by adopting an entropy weight method, and taking the weight as a coefficient for evaluating the health state; building a health assessment model; and carrying out weighted average calculation by adopting a fuzzy operator. The method has the advantages of being real-time, high in robustness, excellent in environmental interference resistance and the like.

Description

Technical Field

[0001] The present invention belongs to the technical field of health monitoring of mechanical rotating equipment, and in particular, relates to a data-driven method and system for health assessment of roller operation status. Background Art

[0002] As an important transmission medium for cargo transshipment, belt conveyor lines play a vital role in various industries due to their high transportation efficiency, simple installation, and easy use. Especially in the coal industry, the mining, transshipment, and fine picking of coal all require the important transportation function of belt conveyor lines. As one of the core accessories on belt conveyor lines, rollers are important accessories that support the sliding transportation of belts. Once a failure occurs, it may cause damage to the belt at the least, or even cause major failures such as belt tearing and burning. Therefore, real-time monitoring of the operating status of rollers has become one of the important technical bottlenecks restricting the intelligent development of belt conveyor lines. Health assessment of roller operating status is one of the core technologies to achieve real-time monitoring.

[0003] There are several key technical problems in the existing technology in the health assessment and real-time monitoring of roller operation status:

[0004] 1. Insufficient real-time monitoring: Traditional roller monitoring methods mostly rely on manual inspections or regular maintenance. This method not only has a time lag and cannot capture minor faults in roller operation in time, but also often cannot detect and take measures in time after abnormal situations occur, resulting in more serious damage or downtime.

[0005] 2. Single data, unable to accurately predict multiple faults: Existing monitoring systems usually rely on simple parameters (such as vibration or temperature) to determine the status of rollers. These single data sources are difficult to fully reflect the health of rollers, especially in complex environments, and cannot effectively identify multiple types of faults. For major faults such as belt tearing and burning, traditional methods are difficult to predict in a timely and accurate manner.

[0006] 3. Environmental interference and data instability: When the roller is running in harsh working conditions such as high temperature and high dust, the data collected by the sensor is easily interfered by environmental noise, resulting in increased errors in the monitoring results. Traditional monitoring technology cannot effectively handle these interference factors, resulting in a decrease in the accuracy of fault identification.

[0007] 4. High maintenance cost: Since existing technologies cannot achieve efficient and accurate online monitoring, enterprises usually need to conduct frequent manual inspections and equipment maintenance, which not only increases labor costs, but also leads to excessive maintenance of equipment when there is no fault, thus wasting resources.

[0008] Therefore, the real-time monitoring of the health status of rollers in the existing technology faces technical bottlenecks such as poor real-time performance, single data, large environmental interference and high maintenance costs, which limits the intelligent development of belt conveyor lines. Summary of the invention

[0009] In view of the problems existing in the prior art, the present invention provides a data-driven roller operating status health assessment method and system, which mainly relies on the intelligent sensing data acquisition system installed on the roller support seat to obtain the vibration data for monitoring the operating status of the roller equipment, and adopts time-frequency synchronization processing methods, filtering processing, time-frequency domain analysis and other processing methods to process the obtained vibration signal and extract multiple feature quantities. On this basis, the information entropy calculation method is adopted to calculate various weight information of multiple feature quantities, and a convolutional neural network is adopted to complete the characterization of the health indicator system of multiple feature quantities. Finally, according to the fuzzy operation of the weight information and the health indicator characterization system, the real-time health assessment result of the roller operating status is obtained.

[0010] The present invention is implemented as follows: a data-driven roller running state health assessment method, comprising:

[0011] S1: During the real-time operation of the roller, a vibration data sensing system is installed near its support frame to complete the collection of required data;

[0012] S2: Sampled signal y(t) obtained for S1 n ,The time-frequency domain feature extraction method is used for feature extraction;

[0013] S3: Based on the eight characteristic signals obtained in S2, a health characteristic indicator system matrix is ​​formed;

[0014] S4: Complete the calculation of all index parameters for the health assessment of the roller operation status and obtain the health index representation;

[0015] S5: According to the indicator parameter system obtained in S4, the entropy weight method is used to calculate the weight of a single indicator in the indicator parameter system as a coefficient for evaluating its health status;

[0016] S6: Build a health assessment model based on the health index representation parameters obtained in S4;

[0017] S7: Based on the deviation distance characteristic root obtained in S6 and the characteristic weight obtained in S5, a fuzzy operator is used to perform weighted average calculation.

[0018] Furthermore, in S1, the collected parameter is recorded as x(t), and the data sampling period is t s , the sampling frequency is f s ;

[0019] According to the continuous time signal x(t), according to the sampling period t s and sampling frequency f s , the signal can be sampled as follows:

[0020] y(t) n =x(t) / t s (1)

[0021] Further, the S2 mainly includes the following:

[0022] (1) There are three characteristics of the statistical domain, as follows:

[0023] Mean:

[0024] [y Mean ] n =mean(y n ) (2)

[0025] Standard Deviation:

[0026] [y std ] n = Std(y n ) (3)

[0027] Variance:

[0028] [y var ] n =Var(y n ) (4)

[0029] (2) Spectral Domain has three characteristics:

[0030] Wavelet Absolute Mean

[0031] [y wam ] n =Wam(y n ) (5)

[0032] Wavelet Absolute Mean

[0033] [y wmean ] n =Wmean(y n ) (6)

[0034] Maximum peak of spectrum

[0035] [y wmax ]n =Wmax(y n ) (7)

[0036] (3) Temporal Domain: Two characteristics:

[0037] Autocorrelation coefficient:

[0038] [y con ] n =conf(y n ,y n ) (8)

[0039] The Number of Maximum Peaks:

[0040] [y ff ] n = ff(y n ) (9)

[0041] Furthermore, the health characteristic indicator system matrix in S3 is as follows:

[0042]

[0043] Further, the S5 calculation method is as follows:

[0044] (1) Set up the index parameter system [T] n =(Z ij ) m*n ,for

[0045]

[0046] (2) Standardize the above indicator parameters to solve the problem of differences between data dimensions. The main methods are as follows:

[0047] Positive indicator normalization:

[0048]

[0049] Negative indicator normalization:

[0050]

[0051] After processing, the new health indicator parameter matrix is ​​obtained as follows:

[0052]

[0053] (3) Calculate the numerical weight P of the i-th item of the j-th indicator ij :

[0054]

[0055] (4) Calculate the k value and the entropy value e of the jth index j :

[0056]

[0057] Where: k>0; e j ≥0

[0058] (5) Calculate the variation index (coefficient of variation), the variation index d of the jth indicator j =1-e j (j=1,2,...,n)

[0059] (6) Calculate the entropy weight coefficient ω j

[0060] The weight ω of the jth indicator j for:

[0061]

[0062] Furthermore, the main implementation contents of S6 are as follows:

[0063] The same sampling method is used to collect the roller operation status data under normal conditions as a reference for health assessment. The matrix mapping method is used to calculate the deviation between the real-time collected data and the benchmark data, as follows:

[0064] Assume that the reference data is u(t), the sampling frequency and sampling period are f respectively. s and t s ;

[0065] Using the contents in steps 2 and 3, obtain the health index characterization parameters under standard data, denoted as [T]' n ', the standard distance is calculated based on the health index characterization parameters under the standard data, as follows:

[0066] d([T]' n ',[T] n )=|{[T]' n '} j -{[T] n} j | (18)

[0067] Where, j = 1, 2, ..., n;

[0068] According to formula (18), we can get a set of matrix calculation formulas, which are set as follows:

[0069] D(ij)=d([T]'n ',[T] n ) (19)

[0070] It can be seen that D(ij) is a matrix with i rows and j columns, that is, each row represents a set of characteristic indicators;

[0071] For the rows of D(ij), take them out separately and record them as [D(ij)] i ; then for [D(ij)] i The matrix is ​​used to find the maximum characteristic root and obtain λ i .

[0072] Further, the S7 is specifically as follows:

[0073] R(i)=[D(ij)] i *ω j (20)

[0074] r=∑R(i) / i (21)

[0075] The health assessment result is obtained, and the result must fall between 0 and 1, and the health level of 0 to 1 is divided as follows:

[0076] 0~0.15: scrapped;

[0077] 0.15~0.45: Fault;

[0078] 0.45~0.75: maintenance;

[0079] 0.75~1: Healthy.

[0080] Another object of the present invention is to provide a data-driven roller running state health assessment system based on the data-driven roller running state health assessment method, the system specifically comprising:

[0081] The data acquisition module is designed to collect the required data by installing a vibration data sensing system near the support frame during the real-time operation of the roller;

[0082] The feature extraction module is connected to the data acquisition module to obtain the sampled signal y(t) from S1. n ,The time-frequency domain feature extraction method is used for feature extraction;

[0083] The matrix composition module is connected with the feature extraction module, and forms a health feature indicator system matrix according to the eight feature signals obtained by S2;

[0084] The characterization acquisition module is connected with the matrix composition module to complete the calculation of all index parameters for the health assessment of the roller operation status and obtain the health index characterization;

[0085] The indicator weight calculation module is connected to the representation acquisition module, and according to the acquired indicator parameter system, the entropy weight method is used to calculate the weight of a single indicator in the indicator parameter system as a coefficient for evaluating its health status;

[0086] The evaluation model building module is connected with the representation acquisition module to build a health evaluation model according to the acquired health index representation parameters;

[0087] The weighted average calculation module is connected to the evaluation model building module. It uses the fuzzy operator to perform weighted average calculation based on the deviation distance feature root and the feature weight obtained by S5.

[0088] Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the data-driven roller operating status health assessment method.

[0089] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the data-driven roller operating status health assessment method.

[0090] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0091] First, the present invention can effectively complete the health assessment of the roller operation status based on data drive, thereby achieving the goal of intelligent monitoring of the roller according to the health assessment level. The present invention has the advantages of real-time, strong robustness, and excellent resistance to environmental interference.

[0092] The technical solution of the present invention solves several key problems in the prior art in industrial applications and has achieved significant technical progress, which is mainly reflected in the following aspects:

[0093] 1. Solve the problem of insufficient vibration monitoring accuracy in existing technologies

[0094] Existing roller monitoring systems usually rely on traditional mechanical methods or simple signal processing, which cannot accurately capture the tiny vibration signals of rollers during operation. This monitoring method is prone to miss early fault signals, resulting in major failures of rollers during long-term operation, such as belt tearing or equipment damage. By introducing a high-precision vibration data sensing system, the present invention can collect vibration signals during roller operation in real time, and combined with time-frequency domain feature extraction technology, it effectively improves the accuracy of capturing abnormal behavior of rollers and realizes early fault detection and early warning.

[0095] 2. Solve the problem of incomplete data processing in existing technologies

[0096] In the prior art, the monitoring and analysis of the roller's operating status is often limited to a single signal feature, and it is impossible to conduct a comprehensive analysis of the roller's health status. The present invention constructs a health feature indicator system and uses multi-dimensional feature signals in the statistical domain, spectral domain, and time domain to form a comprehensive health assessment method. This multi-signal comprehensive assessment method can effectively avoid the errors caused by the fluctuation of a single signal and provide a more comprehensive and accurate health status assessment result.

[0097] 3. Solve the problem of lack of intelligent weight calculation in existing technology

[0098] In the existing roller monitoring system, usually no intelligent algorithm is used to assign weights to various characteristic signals, resulting in inaccurate evaluation results. By adopting the entropy weight method, the present invention can scientifically calculate the weights of each characteristic quantity and objectively measure the importance of each signal in the health status assessment. This automated weight assignment not only improves the accuracy of the algorithm, but also enables the system to have adaptive capabilities, and can automatically adjust the parameters of the evaluation model according to actual data, thereby improving the accuracy and robustness of the evaluation.

[0099] 4. Solve the problem of delayed fault assessment and early warning in existing technologies

[0100] The roller monitoring system in the prior art usually relies on offline detection and cannot perform health assessment and early warning in real time. It is often processed only after the roller has seriously failed, resulting in equipment shutdown and reduced production efficiency. The present invention uses real-time monitoring and data-driven health assessment models to instantly evaluate the health status of the roller during operation, and combines fuzzy operator weighted calculation to achieve intelligent fault early warning. When the assessment result is close to the fault threshold, the system will automatically issue a warning signal, effectively reducing the risk of failure, reducing equipment maintenance costs, and ensuring production continuity.

[0101] Accurate monitoring: The monitoring accuracy of the roller operating status is significantly improved by introducing the vibration data perception system and time-frequency domain feature extraction technology.

[0102] Comprehensive analysis: The construction of a multi-dimensional health characteristic indicator system avoids the limitations of a single signal and provides a more comprehensive health status assessment.

[0103] Intelligent assessment: By calculating weights through the entropy weight method, an intelligent health assessment model is constructed to improve the adaptability and accuracy of the system.

[0104] Real-time warning: Based on the data-driven fault warning mechanism, real-time detection and warning of roller faults are achieved, reducing the risk of equipment downtime.

[0105] These technological advances have enabled the present invention to significantly improve the reliability, operating efficiency and fault prevention capabilities of equipment in the industrial application of roller operation status monitoring, providing strong technical support for modern production.

[0106] Second, the present invention solves the problems of insufficient monitoring accuracy and real-time performance in the prior art through a data-driven roller operating status health assessment method. Traditional roller monitoring methods mostly rely on regular manual inspections or monitoring methods based on simple parameters, which cannot achieve continuous monitoring of equipment and early warning of faults. The present invention extracts time-frequency domain features of vibration data and combines it with the construction of a health feature indicator system to reflect the operating status of rollers in real time and accurately, avoiding the lag of manual detection and improving the safety and operating efficiency of conveying equipment.

[0107] The existing technology is often affected by environmental interference under complex working conditions, resulting in inaccurate monitoring data, which in turn affects the equipment status assessment. The present invention introduces the entropy weight method to calculate the weights of various monitoring indicators, effectively solving the noise interference problem of various signals in complex environments, and ensuring the robustness and adaptability of the assessment model. This method not only improves the accuracy of roller health assessment, but also can adapt to adverse environments such as high temperature and high dust, significantly improving the reliability of the system.

[0108] In terms of multi-fault identification, traditional methods are mostly limited to single-type fault identification and are difficult to deal with complex multi-fault scenarios. The present invention successfully constructs a multi-fault classification model by combining support vector machines and fuzzy operators, which can effectively distinguish fault types under different states and accurately predict possible abnormalities of the equipment. This advancement solves the technical bottleneck of the existing technology in multi-fault processing and significantly improves the breadth and accuracy of roller fault identification.

[0109] In short, the technical progress of the present invention lies in the realization of intelligent and data-driven health assessment of the roller operation status, which greatly improves the monitoring efficiency of the equipment operation status, reduces maintenance costs, and demonstrates significant application value and economic benefits in various industrial application scenarios. This breakthrough not only optimizes the operation and maintenance of industrial conveying systems, but also provides technical support for future intelligent equipment monitoring and management. BRIEF DESCRIPTION OF THE DRAWINGS

[0110] Figure 1 It is a flow chart of a data-driven roller running status health assessment method provided by an embodiment of the present invention;

[0111] Figure 2 It is a structural diagram of a data-driven roller running status health assessment system provided by an embodiment of the present invention;

[0112] In the figure: 1. Data acquisition module; 2. Feature extraction module; 3. Matrix composition module; 4. Representation acquisition module; 5. Index weight calculation module; 6. Evaluation model building module; 7. Weighted average calculation module. DETAILED DESCRIPTION

[0113] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0114] like Figure 1 As shown, an embodiment of the present invention provides a data-driven method for evaluating the health of a roller operation state, comprising:

[0115] S1: During the real-time operation of the roller, a vibration data sensing system is installed near its support frame to complete the collection of required data;

[0116] S2: Sampled signal y(t) obtained for S1 n ,The time-frequency domain feature extraction method is used for feature extraction;

[0117] S3: Based on the eight characteristic signals obtained in S2, a health characteristic indicator system matrix is ​​formed;

[0118] S4: Complete the calculation of all index parameters for the health assessment of the roller operation status and obtain the health index representation;

[0119] S5: According to the indicator parameter system obtained in S4, the entropy weight method is used to calculate the weight of a single indicator in the indicator parameter system as a coefficient for evaluating its health status;

[0120] S6: Build a health assessment model based on the health index representation parameters obtained in S4;

[0121] S7: Based on the deviation distance characteristic root obtained in S6 and the characteristic weight obtained in S5, a fuzzy operator is used to perform weighted average calculation.

[0122] The data-driven roller running status health assessment method provided by the embodiment of the present invention implements a comprehensive assessment of the health status of the roller through the following steps:

[0123] 1. Data collection and perception system deployment (S1):

[0124] A vibration data sensing system is installed near the roller support frame to obtain real-time vibration data of the roller during operation. The sensing system includes multiple sensors, such as accelerometers, which can effectively monitor the vibration signals of the roller during operation. These signals contain important information related to the mechanical health of the roller, such as vibration amplitude, frequency changes, etc., which are the basic data for subsequent evaluation of the roller status.

[0125] 2. Feature extraction (S2):

[0126] For the collected vibration data, the time-frequency domain feature extraction method is used. This method combines the analysis of the time domain and the frequency domain to fully capture the dynamic characteristics of the signal. For example, the time domain features include the mean, variance, skewness, etc. of the signal, while the frequency domain features extract the signal's spectrum energy, main frequency and other information. Through this feature extraction method, eight characteristic signals of the roller's operating status can be effectively extracted to fully characterize the roller's health status.

[0127] 3. Construction of health characteristic indicator system matrix (S3):

[0128] The eight extracted characteristic signals constitute a health characteristic index system matrix. This matrix can be regarded as the "health file" of the roller's operating status, which reflects the detailed status of the roller's operation under different characteristic signals. Each column in the matrix represents a characteristic signal, and each row represents the roller's status in a time segment, ultimately forming a time series characteristic diagram of the roller's status.

[0129] 4. Calculation of health index (S4):

[0130] According to the constructed health characteristic index system, various indicators are calculated and health index is generated. These indexes reflect the overall health level of the roller by quantifying the operating status characteristics of the roller. For example, the health index can be obtained by standardizing and weighted averaging various characteristic values. The higher the value, the better the operating status of the roller, and vice versa, it indicates the potential risk of failure.

[0131] 5. Calculation of indicator weights (S5):

[0132] In order to improve the accuracy of health assessment, the entropy weight method is used to calculate the weight of each characteristic signal in the health assessment. The entropy weight method automatically assigns different weights to each characteristic according to the volatility and importance of each characteristic signal. Through this method, it is possible to more objectively determine which signals have a greater impact on the health status of the roller, thereby enhancing the accuracy of the assessment results.

[0133] 6. Construction of health assessment model (S6):

[0134] Using the health index and weight information obtained from the above calculations, a roller health status assessment model is built. This model comprehensively inputs various health parameters of the roller, and generates the health assessment results of the roller through mathematical models and algorithm calculations. This model can effectively reflect the operating status of the roller and provide support for decision-making.

[0135] 7. Weighted fuzzy evaluation (S7):

[0136] In the final evaluation stage, fuzzy operators are used to perform weighted average calculations based on the deviation distance characteristic root and the previously obtained indicator weights. Fuzzy operators can process complex and fuzzy information, and ultimately generate roller health scores or fault warning signals by weighted summaries of various characteristic values. This process can further improve the accuracy of the evaluation and support intelligent prediction and decision-making of roller status.

[0137] Through the above steps, the present invention realizes a comprehensive evaluation of the operating status of the roller. By using a data-driven approach, through feature extraction, weight allocation, health index calculation and fuzzy evaluation model, the health status of the roller can be evaluated in real time and accurately, providing important support for the maintenance and fault prevention of industrial equipment.

[0138] The working principle of the roller operation health assessment method of the present invention is based on the collection and analysis of vibration data, and the real-time monitoring and assessment of the roller health status is achieved in a data-driven manner. First, a vibration data sensing system is installed near the support frame of the roller to collect the operation data of the roller in real time. These data mainly include the vibration signals generated by the roller during operation. The collected raw signal data is processed in the time domain and frequency domain, which can provide basic data support for subsequent feature extraction.

[0139] Secondly, the collected vibration signals are processed using the time-frequency domain feature extraction method to extract a variety of characteristic signals. This step mainly calculates the characteristic quantities in the statistical domain, spectrum domain and time domain, such as mean, standard deviation, wavelet absolute mean, autocorrelation coefficient, etc. These characteristic signals can fully describe the vibration behavior of the roller during operation. By analyzing these signals, a health characteristic indicator system can be constructed to lay the foundation for subsequent health status assessment.

[0140] The third step is to calculate the weight of the obtained characteristic signals, and calculate the weight of each characteristic signal through the entropy weight method to obtain the contribution of each characteristic signal to the health status of the roller. The entropy weight method can effectively measure the importance of each characteristic quantity, making the evaluation results more scientific and accurate. Through these weight coefficients, a health index characterization system can be constructed, which can be used for the subsequent health assessment model construction.

[0141] Finally, after obtaining the health index characterization system, a health assessment model is constructed, and the weights of each feature quantity are combined to perform weighted average calculation through fuzzy operators. Through this model, the health status of the roller can be comprehensively evaluated. The evaluation result quantifies the health status into a numerical value, and according to the set health level threshold, it is judged whether the roller is in a healthy, maintenance or fault state, realizing real-time monitoring and early warning of the roller operation status.

[0142] In S1, the collected parameter is recorded as x(t), and the data sampling period is t s , the sampling frequency is f s ;

[0143] According to the continuous time signal x(t), according to the sampling period t s and sampling frequency f s , the signal can be sampled as follows:

[0144] y(t) n =x(t) / t s (1)

[0145] The S2 mainly includes the following:

[0146] (1) There are three characteristics of the statistical domain, as follows: Mean:

[0147] [y Mean ] n =mean(y n ) (2) Standard Deviation:

[0148] [y std ] n = Std(y n ) (3) Variance

[0149] [y var ] n =Var(y n ) (4)

[0150] (2) Spectral Domain has three characteristics:

[0151] Wavelet Absolute Mean

[0152] [y wam ] n =Wam(y n ) (5) Wavelet Absolute Mean

[0153] [y wmean ] n =Wmean(y n ) (6) Maximum peak value of the spectrum

[0154] [y wmax ] n =Wmax(y n ) (7)

[0155] (3) Temporal Domain: Two characteristics:

[0156] Autocorrelation coefficient:

[0157] [y con ] n =conf(y n ,y n ) (8) The Number of Maximum Peaks:

[0158] [y ff ] n = ff(y n ) (9) The health characteristic indicator system matrix in S3 is as follows:

[0159]

[0160] The S5 calculation method is as follows:

[0161] (1) Set up the index parameter system [T] n =(Z ij ) m*n ,for

[0162]

[0163] (2) Standardize the above indicator parameters to solve the problem of differences between data dimensions. The main methods are as follows:

[0164] Positive indicator normalization:

[0165]

[0166] Negative indicator normalization:

[0167]

[0168] After processing, the new health indicator parameter matrix is ​​obtained as follows:

[0169]

[0170] (3) Calculate the numerical weight P of the i-th item of the j-th indicator ij :

[0171]

[0172] (4) Calculate the k value and the entropy value e of the jth index j :

[0173]

[0174] Where: k>0; e j ≥0

[0175] (5) Calculate the variation index (coefficient of variation), the variation index d of the jth indicator j =1-e j (j=1,2,...,n)

[0176] (6) Calculate the entropy weight coefficient ω j

[0177] The weight ω of the jth indicator j for:

[0178]

[0179] The main implementation contents of S6 are as follows:

[0180] The same sampling method is used to collect the roller operation status data under normal conditions as a reference for health assessment. The matrix mapping method is used to calculate the deviation between the real-time collected data and the benchmark data, as follows:

[0181] Assume that the reference data is u(t), the sampling frequency and sampling period are f respectively. s and t s ;

[0182] Using the contents in steps 2 and 3, obtain the health index characterization parameters under standard data, denoted as [T]' n ', the standard distance is calculated based on the health index characterization parameters under the standard data, as follows:

[0183] d([T]' n ',[T] n )=|{[T]' n '} j -{[T] n} j | (18)

[0184] Where, j = 1, 2, ..., n;

[0185] According to formula (18), we can get a set of matrix calculation formulas, which are set as follows:

[0186] D(ij)=d([T]' n ',[T] n ) (19)

[0187] It can be seen that D(ij) is a matrix with i rows and j columns, that is, each row represents a set of characteristic indicators;

[0188] For the rows of D(ij), take them out separately and record them as [D(ij)] i ; then for [D(ij)] i The matrix is ​​used to find the maximum characteristic root and obtain λ i .

[0189] The S7 is specifically as follows:

[0190] R(i)=[D(ij)] i *ω j (20)

[0191] r=∑R(i) / i (21)

[0192] The health assessment result is obtained, and the result must fall between 0 and 1, and the health level of 0 to 1 is divided as follows:

[0193] 0~0.15: scrapped;

[0194] 0.15~0.45: Fault;

[0195] 0.45~0.75: maintenance;

[0196] 0.75~1: Healthy.

[0197] like Figure 2 As shown, an embodiment of the present invention provides a data-driven roller running state health assessment system based on the data-driven roller running state health assessment method, and the system specifically includes:

[0198] Data acquisition module 1, during the real-time operation of the roller, a vibration data sensing system is installed near its support frame to complete the collection of required data;

[0199] Feature extraction module 2, connected to data acquisition module 1, obtains the sampled signal y(t) from S1 n ,The time-frequency domain feature extraction method is used for feature extraction;

[0200] The matrix composition module 3 is connected to the feature extraction module 2, and forms a health feature index system matrix according to the eight feature signals obtained by S2;

[0201] The characterization acquisition module 4 is connected to the matrix composition module 3 to complete the calculation of all index parameters for the health assessment of the roller operation status and obtain the health index characterization;

[0202] The indicator weight calculation module 5 is connected to the representation acquisition module 4, and uses the entropy weight method to calculate the weight of a single indicator in the indicator parameter system according to the acquired indicator parameter system as a coefficient for evaluating its health status;

[0203] The evaluation model building module 6 is connected to the representation acquisition module 4 and builds a health evaluation model according to the acquired health index representation parameters;

[0204] The weighted average calculation module 7 is connected to the evaluation model building module 6, and performs weighted average calculation using a fuzzy operator based on the deviation distance feature root and the feature weight obtained in S5.

[0205] An embodiment of the present invention provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the data-driven roller running status health assessment method.

[0206] An embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the data-driven roller running status health assessment method.

[0207] The following are two specific embodiments of the data-driven roller operation health assessment method:

[0208] Embodiment 1:

[0209] Health Assessment of Rollers in Mine Conveyor Systems

[0210] 1. Data collection system deployment: Install vibration sensors near the roller support frame of the mine conveyor belt system to obtain real-time vibration data of the roller during operation. Collect data once a second and transmit it to the central monitoring system via a wireless network.

[0211] 2. Time-frequency domain feature extraction: Through the collected vibration signal, the time-frequency domain feature extraction method is applied to extract the time domain features (such as mean, variance) and frequency domain features (such as spectrum energy, main frequency) of the vibration signal. A total of eight feature signals are extracted.

[0212] 3. Calculation of health index: The health index of each roller is calculated by forming a health characteristic index system matrix based on the time-frequency domain characteristics. The weight of each characteristic signal is assigned through the entropy weight method to generate a comprehensive health status index for each roller.

[0213] 4. Application of fuzzy assessment model: Fuzzy operators are used to calculate the weighted average of the deviation distance and health weight of the roller, and finally the health assessment result of the roller is obtained. The results are displayed through visualization tools, and abnormal conditions are warned in real time to prompt equipment maintenance personnel to conduct inspections.

[0214] Embodiment 2:

[0215] Abnormal detection of rollers in port conveying system

[0216] 1. Data collection and system configuration: In the bulk cargo conveying system of the port, vibration sensors are installed at both ends of the roller bracket. Each sensor collects the operating vibration data of the roller at a sampling frequency of 1 Hz and transmits it to the monitoring center via wired transmission.

[0217] 2. Feature extraction and health index system construction: Perform time-frequency domain analysis on the collected vibration data to extract multiple characteristic signals including vibration amplitude, frequency and acceleration. The eight characteristic signals are combined into a health index system matrix and standardized.

[0218] 3. Health assessment and weight calculation: By calculating the health index of eight characteristic signals, the system uses the entropy weight method to calculate the weight of each characteristic signal, and calculates the health status score of the roller based on the characteristic weight and deviation distance.

[0219] 4. Result analysis and early warning: The system automatically generates a health assessment report by using fuzzy operators to perform weighted calculations on the health status of rollers. Early warnings are issued for rollers with low health scores, prompting equipment maintenance personnel to conduct troubleshooting and troubleshooting in advance.

[0220] These two examples demonstrate the application of this method in mine and port conveying systems. Through data-driven vibration analysis, weight distribution and fuzzy evaluation model, health monitoring and fault warning of the roller operating status can be achieved.

[0221] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A data-driven method for evaluating the health of roller operation status, characterized in that: include: S1: During the real-time operation of the roller, a vibration data sensing system is installed near its support frame to complete the collection of required data; S2: Sampled signal y(t) obtained for S1 n ,The time-frequency domain feature extraction method is used for feature extraction; S3: Based on the eight characteristic signals obtained in S2, a health characteristic indicator system matrix is ​​formed; S4: Complete the calculation of all index parameters for the health assessment of the roller operation status and obtain the health index representation; S5: According to the indicator parameter system obtained in S4, the entropy weight method is used to calculate the weight of a single indicator in the indicator parameter system as a coefficient for evaluating its health status; S6: Build a health assessment model based on the health index representation parameters obtained in S4; S7: Based on the deviation distance characteristic root obtained in S6 and the characteristic weight obtained in S5, a fuzzy operator is used to perform weighted average calculation.

2. The data-driven roller running status health assessment method according to claim 1 is characterized in that: In S1, the collected parameter is recorded as x(t), and the data sampling period is t s , the sampling frequency is f s ; According to the continuous time signal x(t), according to the sampling period t s and sampling frequency f s , the signal can be sampled as follows: y(t) n =x(t) / t s (1) 3. The data-driven roller running status health assessment method according to claim 1 is characterized in that: The S2 mainly includes the following: (1) There are three characteristics of the statistical domain, as follows: Mean: [y Mean ] n =mean(y n ) (2) Standard Deviation: [y std ] n =Std(y n ) (3) Variance: [and var ] n =Var(y n ) (4) (2) Spectral Domain has three characteristics: Wavelet Absolute Mean [and wam ] n =Wam(y n ) (5) Wavelet Absolute Mean [the wmean ] n =Wmean(y n ) (6) Maximum peak of spectrum [and wmax ] n =Wmax(y n ) (7) (3) Temporal Domain: Two characteristics: Autocorrelation coefficient: [and con ] n =conf(and n ,and n ) (8) The Number of Maximum Peaks: [the ff ] n =ff(y n ) (9) 4. The data-driven roller running status health assessment method according to claim 1, characterized in that: The health characteristic indicator system matrix in S3 is as follows:

5. The data-driven roller running status health assessment method according to claim 1, characterized in that: The S5 calculation method is as follows: (1) Set up the index parameter system [T] n =(Z ij ) m*n ,for (2) Standardize the above indicator parameters to solve the problem of differences between data dimensions. The main methods are as follows: Positive indicator normalization: Negative indicator normalization: After processing, the new health indicator parameter matrix is ​​obtained as follows: (3) Calculate the numerical weight P of the i-th item of the j-th indicator ij : (4) Calculate the k value and the entropy value e of the jth index j : Where: k>0; e j ≥0 (5) Calculate the variation index (coefficient of variation), the variation index d of the jth indicator j =1-e j (j=1,2,...,n) (6) Calculate the entropy weight coefficient ω j The weight ω of the jth indicator j for:

6. The data-driven roller running status health assessment method according to claim 1, characterized in that: The main implementation contents of S6 are as follows: The same sampling method is used to collect the roller operation status data under normal conditions as a reference for health assessment. The matrix mapping method is used to calculate the deviation between the real-time collected data and the benchmark data, as follows: Assume that the reference data is u(t), the sampling frequency and sampling period are f respectively. s and t s ; Using the contents in steps 2 and 3, obtain the health index characterization parameters under standard data, denoted as [T]' n ', the standard distance is calculated based on the health index characterization parameters under the standard data, as follows: d([T]' n ',[T] n )=|{[T]' n '} j -{[T] n } j | (18) Where, j = 1, 2, ..., n; According to formula (18), we can get a set of matrix calculation formulas, which are set as follows: D(ij)=d([T]' n ',[T] n ) (19) It can be seen that D(ij) is a matrix with i rows and j columns, that is, each row represents a set of characteristic indicators; For the rows of D(ij), take them out separately and record them as [D(ij)] i ; then for [D(ij)] i The matrix is ​​used to find the maximum characteristic root and obtain λ i .

7. The data-driven roller running status health assessment method according to claim 1, characterized in that: The S7 is specifically as follows: R(i)=[D(ij)] i *ω j (20) r=∑R(i) / i (21) The health assessment result is obtained, and the result must fall between 0 and 1, and the health level of 0 to 1 is divided as follows: 0~0.15: scrapped; 0.15~0.45: Fault; 0.45~0.75: maintenance; 0.75~1: Healthy.

8. A data-driven roller running state health assessment system based on the data-driven roller running state health assessment method according to claims 1-7, characterized in that: The system specifically includes: The data acquisition module is designed to collect the required data by installing a vibration data sensing system near the support frame during the real-time operation of the roller; The feature extraction module is connected to the data acquisition module to obtain the sampled signal y(t) from S1. n ,The time-frequency domain feature extraction method is used for feature extraction; The matrix composition module is connected with the feature extraction module, and forms a health feature indicator system matrix according to the eight feature signals obtained by S2; The characterization acquisition module is connected with the matrix composition module to complete the calculation of all index parameters for the health assessment of the roller operation status and obtain the health index characterization; The indicator weight calculation module is connected to the representation acquisition module, and according to the acquired indicator parameter system, the entropy weight method is used to calculate the weight of a single indicator in the indicator parameter system as a coefficient for evaluating its health status; The evaluation model building module is connected with the representation acquisition module to build a health evaluation model according to the acquired health index representation parameters; The weighted average calculation module is connected to the evaluation model building module. It uses the fuzzy operator to perform weighted average calculation based on the deviation distance feature root and the feature weight obtained by S5.

9. A computer device, characterized in that: The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the data-driven roller running status health assessment method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the data-driven roller running status health assessment method as described in any one of claims 1 to 7.

Images